instabilities of internal gravity wave beams arxiv:1702.07762v2 … · 2017-03-03 · instabilities...
TRANSCRIPT
Instabilities of InternalGravity Wave Beams
Thierry Dauxois, Sylvain Joubaud, PhilippeOdier and Antoine VenailleUniv Lyon, ENS de Lyon, Univ Claude Bernard, CNRS, Laboratoire de
Physique, F-69342 Lyon, France; email: [email protected]
Annu. Rev. Fluid Mech. 2018. 50:1–28
This article’s doi:
10.1146/annurev-fluid-122316-044539
Copyright c© 2018 by Annual Reviews.
All rights reserved
Keywords
internal waves, instability, mean-flow
Abstract
Internal gravity waves play a primary role in geophysical fluids: they
contribute significantly to mixing in the ocean and they redistribute
energy and momentum in the middle atmosphere. Until recently, most
studies were focused on plane wave solutions. However, these solutions
are not a satisfactory description of most geophysical manifestations
of internal gravity waves, and it is now recognized that internal wave
beams with a confined profile are ubiquitous in the geophysical context.
We will discuss the reason for the ubiquity of wave beams in stratified
fluids, related to the fact that they are solutions of the nonlinear gov-
erning equations. We will focus more specifically on situations with a
constant buoyancy frequency. Moreover, in light of recent experimental
and analytical studies of internal gravity beams, it is timely to discuss
the two main mechanisms of instability for those beams. i) The Triadic
Resonant Instability generating two secondary wave beams. ii) The
streaming instability corresponding to the spontaneous generation of a
mean flow.
March 3, 2017
1
arX
iv:1
702.
0776
2v2
[ph
ysic
s.fl
u-dy
n] 2
Mar
201
7
Contents
1. INTRODUCTION .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22. THE DYNAMICS OF STRATIFIED FLUIDS AND ITS SOLUTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Basic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32.2. Linear Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.3. Nonlinear Terms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3. TRIADIC RESONANT INSTABILITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73.2. The simplest case of Plane Waves Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83.3. Why does the Finite Width of Internal Waves Beam Matter? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.4. Energy Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113.5. Theory in the Nearly Inviscid Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.6. Effect of a mean advective flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133.7. Effect of the rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4. STREAMING INSTABILITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164.2. Experimental Observations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174.3. Analytical Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194.4. Forcing of Oceanic Mean Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5. CONCLUSIONS AND FUTURE DIRECTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1. INTRODUCTION
Internal gravity waves play a primary role in geophysical fluids (Sutherland 2010): they
contribute significantly to mixing in the ocean (Wunsch & Ferrari 2004) and they redis-
tribute energy and momentum in the middle atmosphere (Fritts & Alexander 2003). The
generation and propagation mechanisms are fairly well understood, as for instance in the
case of oceanic tidal flows (Garrett & Kunze 2007). By contrast, the dissipation mecha-
nisms, together with the understanding of observed energy spectra resulting from nonlinear
interactions between those waves, are still debated (Johnston et al. 2003; MacKinnon &
Winters 2005; Rainville & Pinkel 2006; Callies, Ferrari & Buhler 2014; Alford et al. 2015;
Sarkar & Scotti 2016). Several routes towards dissipation have been identified, from wave-
mean flow interactions to cascade processes, but this remains a fairly open subject from
both theoretical (Craik 1988; Nazarenko 2011) and experimental points of view (Staquet &
Sommeria 2002). The objective of this review is to present important recent progress that
sheds new light on the nonlinear destabilization of internal wave beams, bridging part of
the gap between our understanding of their generation mechanisms based mostly on linear
analysis, and their subsequent evolution through nonlinear effects.
Until recently, most studies were focused on plane wave solutions, which are introduced
in classical textbooks (Gill 1982). Strikingly, such plane waves are not only solutions of the
linearized dynamics, but also of the nonlinear equations (McEwan 1973; Tabaei & Akylas
2003). However, spatially and temporally monochromatic internal wave trains are not a
satisfactory description of most geophysical internal gravity waves (Sutherland 2013). In-
deed, oceanic field observations have rather reported internal gravity beams with a confined
profile (Lien & Gregg. 2001; Cole et al. 2009; Johnston et al. 2011). In the atmosphere,
gravity waves due to thunderstorms also often form beam-like structures (Alexander 2003).
2 Dauxois et al.
Oceanic wave beams arise from the interaction of the barotropic tide with sea-floor topog-
raphy, as has been recently studied theoretically and numerically (Khatiwala 2003; Lamb
2004; Mauge & Gerkema 2008), taking into account transient, finite-depth and nonlin-
ear effects, ignored in the earlier seminal work by Bell (1975). The importance of those
beams has also been emphasized recently in quantitative laboratory experiments (Gostiaux
& Dauxois 2007; King, Zhang & Swinney 2009; Peacock, Echeverri & Balmforth 2008).
From these different works, it is now recognized that internal wave beams are ubiquitous in
the geophysical context.
The interest for internal gravity beams resonates with the usual pedagogical introduc-
tion to internal waves, the Saint Andrew’s cross, which comprises four beams generated by
oscillating a cylinder in a stratified fluid (Mowbray & Rarity 1967). Thorough studies of in-
ternal wave beams can be found in Voisin (2003). Moreover, Tabaei & Akylas (2003) have
realized that an inviscid uniformly stratified Boussinesq fluid supports time-harmonic plane
waves invariant in one transverse horizontal direction, propagating along a direction deter-
mined by the frequency (and the medium through the buoyancy frequency), with a general
spatial profile in the cross-beam direction. These wave beams are not only fundamental
to the linearized dynamics but, like sinusoidal wavetrains, happen to be exact solutions of
the nonlinear governing equations. Remarkably, Tabaei & Akylas (2003) showed that the
steady-state similarity linear solution for a viscous beam (Thomas & Stevenson 1972) is
also valid in the nonlinear regime. In light of the recent experimental and analytical studies
of those internal gravity wave beams, it is thus timely to study their stability properties.
The structure of the review is the following. First, in section 2, we introduce the subject
by presenting concepts, governing equations and approximations that lead to the description
of gravity waves in stratified fluids. We dedicate a special emphasis on the peculiar role
of nonlinearities to explain why internal gravity wave beams are ubiquitous solutions in
oceans and middle atmospheres. Then, in section 3, we discuss the classic Triadic Resonant
Instability that corresponds to the destabilization of a primary wave with the spontaneous
emission of two secondary waves, of lower frequencies and different wave vectors. In addition
to the simple case of plane waves, we discuss in detail the generalization to wave beams
with a finite width. Section 4 is dedicated to the streaming instability, the second important
mechanism for the instability of internal gravity waves beams through the generation of a
mean flow. Finally, in section 5, we draw some conclusions and discuss main future issues.
2. THE DYNAMICS OF STRATIFIED FLUIDS AND ITS SOLUTIONS
2.1. Basic Equations
Let us consider an incompressible non rotating stratified Boussinesq fluid in Cartesian
coordinates (ex,ey,ez) where ez is the direction opposite to gravity. The Boussinesq ap-
proximation amounts to neglecting density variations with respect to a constant refer-
ence density ρref , except when those variations are associated with the gravity term g.
The relevant field to describe the effect of density variations is then the buoyancy field
btot = g (ρref − ρ) /ρref , with ρ(r, t) the full density field, r=(x,y,z) the space coordinates
and t the time coordinate. Let us call ρ0(z) the density of the flow at rest, with buoyancy fre-
quency N(z) = (−g (∂zρ0) /ρref)1/2. The corresponding buoyancy profile g (ρref − ρ0) /ρref
is denoted b0. The buoyancy frequency N varies in principle with the depth z. In the ocean,
N is rather large in the thermocline and weaker in the abyss. For the sake of simplicity,
however, N will be taken constant in the remainder of the paper. In some studies, to ease
www.annualreviews.org • Instabilities of internal wave beams 3
greatly the theoretical analysis, this approximation that looks drastic at first sight can be
relaxed when N changes smoothly by relying on the WKB approximation.
The equations of motion can be written as a dynamical system for the perturbed buoy-
ancy field b = btot − b0 and the three components of the velocity field u = (ux,uy,uz):
∇ · u = 0, (1)
∂tu + u · ∇u = − 1
ρref∇p+ bez + ν∇2u, (2)
∂tb+ u · ∇b+ uzN2 = 0. (3)
with p(r, t) the pressure variation with respect to the hydrostatic equilibrium pressure
P0(z) = P0(0) −∫ z0ρ0(z′)gdz′, and ν the kinematic viscosity. We have neglected the
molecular diffusivity, which would imply a termD∇2b in the right-hand side of Equation (3),
with D the diffusion coefficient of the stratifying element (molecular diffusivity for salt,
thermal diffusivity for temperature). The importance of the dissipative terms with respect
to the nonlinear ones are described by the Reynolds UL/ν and the Peclet numbers UL/D,
with U and L typical velocity and length scales, or equivalently by the Reynolds number
and the Schmidt number ν/D. In many geophysical situations, both Reynolds and Peclet
numbers are large, and molecular effects can be neglected at lowest order. In such cases,
the results do not depend on the Schmidt number. In laboratory settings, the Peclet is
often also very large, at least when the stratification agent is salt, in which case D ≈ 10−9
m2·s−1. However, the viscosity of water is ν ≈ 10−6 m2·s−1, and the corresponding Reynolds
numbers are such that viscous effects can play an important role, as we will see later.
Let us first consider the simplest case of two-dimensional flow, which is invariant in the
transverse y-direction. The non-divergent two-dimensional velocity field is then conveniently
expressed in terms of a streamfunction ψ(x, z) as u = (∂zψ, 0,−∂xψ). Introducing the
Jacobian J(ψ, b) = ∂xψ ∂zb − ∂xb ∂zψ, the dynamical system (1), (2) and (3) is expressed
as
∂t∇2ψ + J(∇2ψ,ψ) = −∂xb+ ν∇4ψ, (4)
∂tb+ J(b, ψ)−N2∂xψ = 0. (5)
Differentiating Equation (4) with respect to time and Equation (5) with respect to the
spatial variable x, and subtracting the latter from the former, one gets finally
∂tt∇2ψ +N2∂xxψ = ν∇4∂tψ + ∂tJ(ψ,∇2ψ) + ∂xJ(b, ψ), (6)
describing the nonlinear dynamics of non-rotating non-diffusive viscous stratified fluids in
two dimensions.
2.2. Linear Approximation
In the linear approximation, assuming vanishing viscosity, the right-hand side of Equa-
tion (6) immediately vanishes leading to the following wave equation for the streamfunction
∂tt∇2ψ +N2∂xxψ = 0. (7)
This equation is striking for several reasons. First, its mathematical structure is clearly dif-
ferent from the traditional d’Alembert equation. Indeed, the spatial differentiation appears
4 Dauxois et al.
at second order in both terms. Time-harmonic plane waves with frequency ω, wave vector
k = (`, 0,m) and wavenumber k = |k| = (`2 + m2)1/2 are solutions of Equation (7), if the
dispersion relation for internal gravity waves
ω = ±N `
k= ±N sin θ, (8)
is satisfied. θ is the angle between wavenumber k and the vertical.
Plane wave:ψ0 e
i(k·r−ωt) + c.c.where c.c. denotes
complex conjugate
The second important remark is that contrary to the usual concentric waves emitted
from the source of excitation when considering the d’Alembert equation, here four differ-
ent directions of propagation are possible depending on the sign of ` and m. This is an
illustration of the anisotropic propagation due to the vertical stratification.
The third remarkable property is that the dispersion relation features the angle of
propagation rather than the wavelength, emphasizing a clear difference between internal
waves and surface waves. This is also a crucial property for this review since it will allow
us to define beams with a general profile, rather than with a single wavenumber.
2.3. Nonlinear Terms
2.3.1. Plane Wave Solutions. It is striking and pretty unusual that plane waves are solu-
tions of the inviscid nonlinear equation (6) even for large amplitudes. Indeed, the stream-
function of the plane wave solution is a Laplacian eigenmode, with ∇2ψ = −k2ψ. Conse-
quently, the first Jacobian term vanishes in Equation (6). Equation (4) leads therefore to
the so-called polarization relation b = −(N2`/ω
)ψ ≡ Pψ, with P the polarization prefac-
tor. Consequently, the second Jacobian in (6) vanishes: J(ψ,Pψ) = 0. To conclude, both
nonlinear terms in Equation (6) vanish for plane wave solutions, that are therefore solutions
of the nonlinear equation, for any amplitude.
2.3.2. Internal Wave Beams. Since the frequency ω is independent of the wavenumber, it
is possible to devise more general solutions, time-harmonic with the same frequency ω, by
superposing several linear solutions associated to the same angle of propagation, but with
different wavenumbers k (McEwan 1973; Tabaei & Akylas 2003). Introducing the along-
beam coordinate ξ = x cos θ − z sin θ, defined along the direction of propagation, and the
cross-beam coordinate η = x sin θ + z cos θ (see Figure 1), the plane wave solution can be
written as
ψ(x, y, z, t) = ψ0 ei(`x+mz−ωt) + c.c. = ψ0 e
ikη e−iωt + c.c. , (9)
since ` = k sin θ and m = k cos θ. If one introduces Q(η) = ikψ0eikη, one obtains the
velocity field u = Q(η)(cos θ, 0,− sin θ)e−iωt + c.c. and the buoyancy perturbation b =
−i(P/k)Q(η)e−iωt + c.c. .
One can actually obtain a wider class of solutions by considering an arbitrary com-
plex amplitude Q(η). Indeed, the fields u and b do not depend on the longitudi-
nal variable ξ. Consequently, after the change of variables, the Jacobians, which read
J(ψ, b) = ∂ξψ ∂ηb− ∂ξb ∂ηψ, simply vanish, making the governing equations linear. As dis-
cussed in Tabaei, Akylas & Lamb (2005), note that uni-directional beams, in which energy
propagates in one direction, involve plane waves with wavenumbers of the same sign only:
Q(η) =∫ +∞0
A(k)eikηdk or Q(η) =∫ 0
−∞A(k)eikηdk.
Internal wave beam:Superposition of
time-harmonic planewaves with anarbitrary profile in
the cross-beam
direction.We will call uniform
beam the specialcase of a internalplane wave with a
confined spatialprofile.
We see that the class of propagating waves that are solutions of the nonlinear dynamics
in a Boussinesq stratified fluid is much more general than plane wave solutions: there
www.annualreviews.org • Instabilities of internal wave beams 5
is a whole family of solutions corresponding to uniform plane waves in the longitudinal
direction ξ, but with a general profile in the cross-beam direction η, as represented in
Figure 1.
(a) (b)
O
cg
cϕ
η
g
θx
ξ
y
z
x
zgη
ξ
u(η)θ
θ
Figure 1
(a) Schematic representation of an internal wave beam and definition of the longitudinal and
cross-beam coordinates ξ and η, of the angle of inclination θ, and finally of the group and phase
velocities cg and cϕ. (b) Geometry of a uniform (along ξ) internal wave beam inclined at an angleθ with respect to the horizontal. The beam profile varies in the cross-beam η direction, and the
associated flow velocity is in the along-beam direction ξ. The transverse horizontal direction is
denoted by y.
Tabaei & Akylas (2003) have generalized those results by computing asymptotic solu-
tions for a slightly viscous nonlinear wave beam with amplitude slowly modulated along ξ
and in time. After considerable manipulation, it turns out that all leading-order nonlinear
advective-acceleration terms in the governing equations of motion vanish, and a uniform
(along ξ) beam, regardless of its profile (along η), represents an exact nonlinear solution
in an unbounded, inviscid, uniformly stratified fluid. This result not only extends the va-
lidity of the Thomas & Stevenson (1972) steady-state similarity solution to the nonlinear
regime, but emphasizes how nonlinearity has only relatively weak consequences. This has
profound and useful outcomes on the applicability of results obtained with linear theory,
for comparisons with field observations, laboratory experiments or numerical simulations.
The vanishing of the nonlinear contributions is really unexpected and results from the
combination of numerous different terms. Tabaei & Akylas (2003) noticed, however, that
the underlying reason for the seemingly miraculous cancellation of the resonant nonlinear
terms was the very same one that had been already pointed out by Dauxois & Young
(1999). After lengthy calculations, in both cases, the reason is a special case of the Jacobi
identity J [A, J(B,C)]+J [C, J(A,B)]+J [J(A,C), B] = 0. Dauxois & Young (1999) were
studying near-critical reflection of a finite amplitude internal wave on a slope to heal the
singularity occurring in the solution of Phillips (1966). Using matched asymptotic, they
took a distinguished limit in which the amplitude of the incident wave, the dissipation, and
the departure from criticality are all small. At the end, although the reconstructed fields do
contain nonlinearly driven second harmonics, they obtained the striking and unusual result
that the final amplitude equation happens to be a linear equation. The underlying reason
was already this Jacobi identity.1
1Studying the mechanism of superharmonic generation, Liang, Zareei & Alam (2017) reportedrecently another situation for which the nonlinear terms vanish in the domain bulk. Interestingly,however, they play a pivotal role through the free surface boundary condition.
6 Dauxois et al.
To conclude, the effects of nonlinearities on plane waves or wavebeams exhibit very
peculiar properties. There are two important points to keep in mind. First, plane waves
and internal wave beams are solutions of the full equation. Second, identifying a solution
does not mean that it is a stable one. This remark is at the core of the present review:
we will focus in the following on the behavior of wave beams with respect to the triadic
resonant and the streaming instabilities.
3. TRIADIC RESONANT INSTABILITY
3.1. Introduction
It was first realized fifty years ago that internal gravity plane waves are unstable to in-
finitesimal perturbations, which grow to form temporal and spatial resonant triads (Davis
& Acrivos 1967; McEwan 1971; Mied 1976). This nonlinear instability produces two
secondary waves that extract energy from a primary one. Energy transfer rates due to this
instability are now well established for plane waves (Staquet & Sommeria 2002).
The instability was observed in several laboratory experiments (Benielli & Sommeria
1998; Clark & Sutherland 2010; Pairaud et al. 2010; Joubaud et al. 2012) and nu-
merical experiments on propagating internal waves (Koudella & Staquet 2006; Wienkers
2015) or reflecting internal tides on a horizontal or sloping boundary (Gerkema, Staquet &
Bouruet-Aubertot 2006; Pairaud et al. 2010; Zhou & Diamessis 2013; Gayen & Sarkar
2013). Oceanic field observations have also confirmed the importance of this instability,
especially close to the critical latitude, where the Coriolis frequency is half of the tidal
frequency (Hibiya, Nagasawa & Niwa 2002; MacKinnon et al. 2013; Sun et al. 2013).
Recent experiments by Bourget et al. (2013), however, followed by a simple model and
numerical simulations by Bourget et al. (2014) as well as a theory by Karimi & Akylas
(2014) have shown that finite-width internal gravity wave beams exhibit a much more
complex behavior than expected in the case of interacting plane waves. This is what will
be discussed in this section.
The Triadic Resonant Instability (TRI) versus the Parametric Subharmonic Instability (PSI)
The classic Triadic Resonant Instability corresponds to the destabilization of a primary wave through the
spontaneous emission of two secondary waves. The frequencies and wave vectors of these three waves are
related by the spatial, k0 = k+ + k−, and the temporal, ω0 = ω+ + ω−, resonance conditions, where the
indices 0 and ± refer respectively to the primary and secondary waves.
In the inviscid case, the most unstable triad corresponds to antiparallel, infinitely long secondary wave
vectors associated with frequencies which are both half of the primary wave frequency: ω+ ' ω− ' ω0/2.
Because of the direct analogy with the parametric oscillator, this particular case defines the Parametric
Subharmonic Instability (PSI). This special case applies to many geophysical situations, and especially for
oceanic applications.
In laboratory experiments, viscosity plays an important role and the two secondary wave frequencies are
different. By abuse of language, some authors have sometimes extended the use of the name PSI to cases
for which secondary waves do not oscillate at half the forcing frequency. To avoid confusion, in the general
case, it is presumably more appropriate to use the acronym TRI.
www.annualreviews.org • Instabilities of internal wave beams 7
3.2. The simplest case of Plane Waves Solutions
3.2.1. Derivation of the Equations and Plane Waves Solutions. Looking for solutions
of the basic equations (4) and (5) as sum of three plane waves as follows b =∑jRj(t)e
i(kj ·r−ωjt) +c.c. and ψ =∑
jΨj(t)e
i(kj ·r−ωjt) +c.c., with j = 0 for the primary
wave and j = ± for the secondary ones, and denoting R the derivative of the amplitude R,
one gets (see for example Hasselman (1967))∑j
[−k2j (Ψj − iωjΨj) + i`jRj − νk4jΨj ]ei(kj ·r−ωjt) + c.c. = −J(∇2ψ,ψ) . (10)
∑j
[Rj − iωjRj − iN2`jΨj ]ei(kj ·r−ωjt) + c.c. = −J(b, ψ) , (11)
The left-hand sides represent the linear parts of the dynamics. Neglecting the nonlinear
terms, as well as the viscous terms and the temporal evolution of the amplitudes, one
recovers the polarization expression Rj = −(N2`j/ωj)Ψj and the dispersion relation ωj =
N |`j |/√`2j +m2
j . This linear system is resonantly forced by the Jacobian nonlinear terms
on the right-hand side when the waves fulfill a spatial resonance condition
k0 = k+ + k− (12)
and a temporal resonance condition
ω0 = ω+ + ω− . (13)
The Jacobian terms in Equations (10) and (11) can then be written as the sum of a resonant
term that will drive the instability, plus some unimportant non resonant terms. Introducing
this result into Equation (10), one obtains three relations between Ψj and Rj for each mode
exp[i(kj · r − ωjt)] with j = 0,+ or −. One gets
R± =1
i`±
[k2±(Ψ± − iω±Ψ±) + νk4±Ψ± + α±Ψ0Ψ∗∓
], (14)
where α± = (`0m∓ − m0`∓)(k20 − k2∓). Here, one traditionally uses the “pump-wave”
approximation, which assumes that over the initial critical growth period of the secondary
waves, the primary wave amplitude, Ψ0, remains constant and that the amplitude varies
slowly with respect to the period of the wave (Ψj ωjΨj). Differentiating the polarization
expression, cumbersome but straightforward calculations (Bourget et al. 2013) lead to first
order to
dΨ±dt
= |I±|Ψ0Ψ∗∓ −ν
2k2±Ψ±, (15)
where I± = (`0m∓ −m0`∓)[ω±(k20 − k2∓) + `±N2(`0/ω0 − `∓/ω∓)]/(2ω±k
2±).
Differentiating Equation (15), one gets
Ψ± = I+I−|Ψ0|2Ψ± −ν2
4k2+k
2−Ψ± −
ν
2(k2+ + k2−)Ψ± . (16)
The general solution is Ψ±(t) = A1,2 exp (σt) +B1,2 exp (σ′t), with σ = −ν(k2+ + k2−)/4 +√(ν/4)2(k2+ − k2−)2 + I+I−|Ψ0|2 and σ′ < 0 < σ.
8 Dauxois et al.
−1−2−3 0 1 2 3+/0
k−k+
k0
m+/m
0
−1
−2
−3
0
1
2
3k+
k−
(a) (b)
σ/max(σ)
k+/k0
(c)
0 2 4
0.4
6
0.6
0.8
8−0.2
0.2
0
1
Resonance locus
k+/k00 2 4 6 8
Viscous caseInviscid case
Figure 2
(a) Resonance locus for the unstable wave vectors (`+,m+) satisfying Equation (17) once the
primary wave vector k0 =(`0,m0) is given. Two examples of vector triads (k0, k+, k−) areshown. The dotted curve is defined by k+ = k0. The solid green curves correspond to the central
branch, while the dashed and dash-dotted black curves correspond to the external branch. (b) and
(c) Corresponding growth rates σ/max(σ) as a function of the normalized wave vectormodulus k+/k0. (b) presents the inviscid case while (c) presents a viscous case corresponding to
Ψ0/ν = 100.
In conclusion, a vanishingly small amplitude noise induces the growth of two secondary
waves by a triadic resonant mechanism. Since their sum gives the primary frequency (see
Equation (13)), ω+ and ω− are subharmonic waves. The growth rate of the instability
depends on the characteristics of the primary wave, namely its wave vector, its frequency
and its amplitude Ψ0, but also on the viscosity ν.
3.2.2. Triads, Resonance Loci and Growth Rates. Using the dispersion relation for internal
waves, the temporal resonance condition leads to (Bourget et al. 2013)
|`0|√`20 +m2
0
=|`+|√`2+ +m2
+
+|`0−`+|√
(`0−`+)2 + (m0−m+)2, (17)
whose solutions are presented in Figure 2a. Once the primary wave vector k0 is defined,
any point of the solid curve corresponds to the tip of the k+ vector, while k− is obtained
by closing the triangle. The choice between the labels + and - is essentially arbitrary and
this leads to the symmetry k → k0 − k in Figure 2a. Without loss of generality, we will
always call k+ the largest wavenumber.
One can observe two distinct parts of this resonance locus, characterized by the position
of k+/k0 with respect to 1. The wavelength of the secondary waves generated by the
instability can be
• both smaller than the primary wavelength: this case corresponds to the external
branch of the resonance locus and implies an energy transfer towards smaller scales
(represented by black curves in Figure 2).
• one larger and the other one smaller: this case corresponds to the central branch of
the resonance locus and implies an energy transfer towards smaller and larger scales
(represented by solid green curves in Figure 2).
Among the different possible solutions on the resonance locus, the one expected to be
seen experimentally or numerically is the one associated with the largest growth rate. In the
inviscid case, the most unstable growth rate occurs for k →∞, with essentially k+ ' −k−,
www.annualreviews.org • Instabilities of internal wave beams 9
and therefore ω+ = ω− = ω0/2. This ultraviolet catastrophe is healed in the presence of
viscosity, which selects a finite wavelength for the maximum growth rate (Hazewinkel &
Winters 2011) as shown in Figure 2c. For typical laboratory scale experiments, the values
of k+ corresponding to significant growth rates are of the same order of magnitude as the
primary wavenumber k0, as can be seen in Figure 2c, with k1/k0 ' 1.5 and k2/k0 ' 2.3.
Thus, TRI corresponds to a direct energy transfer from the primary wave to small scales
where viscous effects come into play, without the need of a turbulent cascade process.
The fact that viscosity has a significant effect on the selection of the excited resonant
triad, preventing any large wave number secondary wave to grow from the instability, has
been observed by Bourget et al. (2013) in laboratory experiments on wave beams. However,
they also found a different type of triads than those predicted by the previous theoretical
arguments. This will be discussed in more detail in the following sections.
3.2.3. Amplitude Threshold for Plane Wave Solutions. The expression for the growth rate
σ implies that the amplitude of the stream function has to be larger than the critical
value |Ψc(`+,m+)| = νk+k−/√
4I+I− to get a strictly positive growth rate (Koudella &
Staquet 2006; Bourget et al. 2013). The threshold for the instability is thus given by
the global minimum of this function of several variables. Let us focus on the particular
case where k+ tends to k0 by considering the following description of the wave vector
components `+ = `0(1 + µ0εα) and m+ = m0(1 + ε) where ε 1, α ≥ 1, while ε and µ0
are positive quantities. Using the dispersion relation, the temporal and spatial resonance
conditions, Bourget et al. (2014) have shown that α = 2 is the only acceptable value to
balance the lowest order terms. Plugging these relations into the expression of I±, one gets
I+ = −`0m0ε+ o(ε) and I− = −`0m0 + o(1), which leads to |Ψc| =√ε νN/(2ω0) + o(ε1/2).
The minimum of this positive expression being zero, it shows that there is no threshold for
an infinitely wide wave beam, even when considering a viscous fluid. Plane wave solutions
are thus always unstable to this Triadic Resonant Instability.
3.3. Why does the Finite Width of Internal Waves Beam Matter?
The above theory for the TRI does not take into account the finite width of the experimental
beam. Qualitatively, the subharmonic waves can only extract energy from the primary wave
if they do not leave the primary beam before they can extract substantial energy (Bourget
et al. 2014). The group velocity of the primary wave is aligned with the beam, but the
group velocity of the secondary waves is definitely not, and these secondary waves eventually
leave the primary wave beam, as illustrated in Figure 3. This is a direct consequence of the
dispersion relation, which relates the direction of propagation to the frequency: a different
frequency, smaller for subharmonic waves, will lead to a shallower angle.
Three comments are in order:
i) The angles between primary and secondary waves strongly influence the interaction
time, and thus the instability.
ii) Secondary waves with small wave vectors, having a larger group velocity cg,± =
(N2 − ω2±)1/2/k±, leave the primary wave beam more rapidly and have less time to grow
in amplitude. Such solutions will therefore be less likely to develop, opening the door to
stabilization of the primary wave by the finite width effect. This clarifies why experiments
with the most unstable secondary waves on the internal branch (small wave vector case) of
the resonance locus (Figure 3) were found to be stable (Bourget et al. 2013) contrary to
10 Dauxois et al.
–30
–20
–10
0
0 15 30 45 0 15 30 45
–30 –15 0 15 30
–30
–20
–10
0
0 15 30 45 0 15 30 45
–30 –15 0 15 30
–30
–20
0 15 30 45 0 15 30 45
cg,−cg,+
cg,0
k−k+k0 = +
(a) (b) (c)
Stable beam Unstable beam
Figure 3
Sketch of the experimental set-up showing the wave generator lying horizontally at the top of the
wave tank with a superimposed snapshot of the vertical density gradient field. (a) The internalwave beam is propagating downward. (b) The instability of the propagating internal wave beam is
visible (Bourget 2014). The tilted dashed rectangle corresponds to the control area for the energy
approach of section 3.4. (c) The vector triad with the three arrows representing the primary wavevector k0 (black) and the two secondary waves vectors k+ (red) and k− (blue). From this triad,
it is possible to deduce the orientation of the group velocities of the three different waves as shownin panels (a) and (b).
the prediction for plane waves. This decisive role of the group velocity of the short-scale
subharmonic waves was identified long ago by McEwan & Plumb (1977).
iii) At the other end of the spectrum, small wavelengths are more affected by dissipation
and will also be less likely to be produced by TRI. Consequently, only a window of secondary
wavelengths is possibly produced by TRI.
3.4. Energy Approach
A simple energy balance proposed by Bourget et al. (2014) makes possible an insightful and
more quantitative estimate for the most unstable triad. We introduce the tilted rectangle
shown in Figure 3 as control area (denoting W the perpendicular beam width) and we
neglect the spatial attenuation of the primary wave in this region (“pump-wave” approxi-
mation). Since secondary waves do not propagate parallel to the primary beam, they exit
the control area from the lateral boundaries without compensation. Equation (15) is thus
modified as follows
dΨ±dt
= |I±|Ψ0Ψ∗∓ −ν
2k2±Ψ± −
|cg,± · ek0 |2W
Ψ± . (18)
The first term represents the interaction with the other plane waves of the triadic resonance,
the second term is due to viscous damping while the third one accounts for the energy leaving
the control area.
One finds here also exponentially growing solutions with a positive growth rate slightly
modified as σ∗ = − (Σ+ + Σ−) /4+√
(Σ+ − Σ−)2 /16 + |I+||I−||Ψ0|2, in which the effective
viscous term now reads Σ± = νk2± + |cg,± · ek0 |/W . The finite width of the beam is
responsible for a new term characterizing the transport of the secondary waves energy
out of the interaction region. For infinitely wide wave beams (W → +∞), one recovers
the growth rate σ obtained in the plane wave case. In contrast, when the beam becomes
narrow (W → 0), the growth rate decreases to zero, leading to a stabilization.
The finite width of a wave beam increases therefore its stability, owing to the transport of
the secondary waves out of the triadic interaction zone of the primary wave beam before they
www.annualreviews.org • Instabilities of internal wave beams 11
can extract substantial energy. This interaction time scales directly with the perpendicular
beam width, W as can be seen from the expression of σ∗.
3.5. Theory in the Nearly Inviscid Limit
A beautiful weakly nonlinear asymptotic analysis of the finite width effect on TRI has been
recently proposed by Karimi & Akylas (2014). Mostly interested by oceanic applications,
they look for subharmonic perturbations in the form of fine-scale, nearly monochromatic
wavepackets with frequency close to one half of the primary frequency; in this limit, usually
called Parametric Subharmonic Instability (see the sidebar distinguishing TRI vs. PSI),
ω± ' ω0/2 = N(sin θ)/2 = N sinφ, that defines the angle φ with the vertical of the wave
vectors k± of opposite directions.
The key ingredient in the analytical derivation is to take advantage of the scale differ-
ence between the width of the primary beam W and the very small carrier wavelength of
the subharmonic wave packets λ± = 2π/k±. The small amplitude expansion is thus char-
acterized by the small parameter µ = λ±/(2πW ). They consider an expansion with not
only the underlying wave beam but also the superimposed subharmonic wavepackets that
appear to order µ. The derivation of the wave-interaction equations leads to six coupled
equations for the two primary beam envelopes (two because of the two phases) and the
four (2 × 2) subharmonic wavepacket envelopes. Fortunately, this system can be reduced
at leading order to only three coupled equations with three unknowns.
Taking a distinguished limit in which not only the amplitude of the primary wave, but
also the nonlinear, dispersive and viscous terms are all small, they obtain a reduced de-
scription of the dynamics. The strategy is as usual to choose the scaling in order to get
comparable magnitudes of the different terms. Interestingly, as only the quadratic interac-
tion is potentially destabilizing for the primary beam, they compare it with the advection
term for subharmonic waves: the former being smaller, this confirms that the resonant
interaction cannot feed the instability in the limited time during which perturbations are
in contact with the underlying beam. Beams with a general profile of finite width are thus
stable to TRI.
Next, they consider the case of beams with profiles in the form of a monochromatic
carrier with O(1) wavelength, modulated by a confined envelope. Functions of the cross-
beam direction η (see Figure 1) and of the appropriate slow time τ , the complex envelopes
Ψ0(η, τ) and Ψ±(η, τ), of the primary and secondary waves are thus generalizations of
the plane wave solutions considered in section 3.2.1. We recover these solutions with an
envelope function independent of the cross-beam coordinate η, while an internal wave beam
(such as the one in Figure 3) will correspond to Ψ0(η) = 1/2 for |η| < 1/2 and zero
otherwise. Introducing the appropriate change of variables and rescaling of the relevant
variables, the beam envelope Ψ0 and the complex subharmonic envelopes Ψ± are linked
through the following three coupled dimensionless equations
∂Ψ±∂τ
= −cg,± · ek0∂Ψ±∂η− νκ2Ψ± + i
Nκ2
ω0sin2 χ |Ψ0|2Ψ± +$Ψ0Ψ∗∓, (19)
∂Ψ0
∂τ= −2$Ψ+Ψ−, (20)
where ν is the renormalized viscous dissipation and κ = 2/µ a rescaled wavenumber mod-
ulus. χ = θ − φ and $ = sinχ cos2 (χ/2) are two geometrical parameters, while τ is
the appropriate slow time for the evolution of the subharmonic wave packets and η the
12 Dauxois et al.
appropriate cross-beam coordinate.
In the appropriate distinguished limit identified by Karimi & Akylas (2014), the non-
linear term is balanced as in section 3.4 by the viscous term, but also by the transport term.
The coupling between the evolution equations occurs through the nonlinear terms, which
allow energy exchange between the underlying beam and subharmonic perturbations. The
subsequent behavior of the complex envelopes Ψ± determines the stability of the beam: if
they are able to extract energy via nonlinear interaction with Ψ0 at a rate exceeding the
speed of linear transport and viscous decay, the beam is unstable.
From this system, it is in principle possible to study the stability of any profile:
1. For example, a time independent beam (Ψ0(η),Ψ± = 0) is a steady state solution
of this system of three equations. The study of its stability relies on looking for the
normal mode solutions Ψ± ∝ exp(στ). For a plane wave, one obtains the growth rate
σ = sinχ cos2 (χ/2) /2− νκ2. A subtle point was carefully emphasized by Karimi &
Akylas (2014). The above expression of σ seems independent of the wave vector dis-
turbance κ in the inviscid limit, but the derivation has extensively used the hypothesis
of fine-scale disturbances, that will of course break down for κ 1. The maximum
growth rate is indeed attained for finite but small κ. Uniform beams (internal plane
waves with a confined envelope) are unstable if the beam is wide enough.
2. Karimi & Akylas (2014) provide also the solution of the initial value problem for
a beam with Ψ0(η) tending towards zero as η tends to infinity. They show the
existence of a minimum value for the unstable perturbation wavenumber κmin =
πcg,±/(2$W∫ +∞−∞ Ψ0(η)dη), corresponding to a maximum wavelength. Therefore,
the possible spatial scale window for secondary wavelengths shrinks towards smaller
scales as the beam is made narrower. Outside this range, no instability is possible
even in the inviscid case.
3. They derive also analytically the minimum width explicitly for the top-hat profile
used in the experiments by Bourget et al. (2013) (Ψ0(η) = 1/2 for |η| < 1/2 and
zero otherwise as shown in Figure 3). They argue that the existence of a minimum
width is valid for a general profile. This minimum is dependent on the beam shape.
To summarize, internal plane waves with a confined envelope are unstable if the beam is
wide enough, while weakly nonlinear beams with a general but confined profile (i.e. without
any dominant carrier wavenumber) are stable to short-scale subharmonic waves.
3.6. Effect of a mean advective flow
Lerisson & Chomaz (2017) have recently studied theoretically and numerically the Triadic
Resonant Instability of an internal gravity beam in presence of a mean advective flow. They
keep constant the local wave vector and wave frequency in the frame moving with the fluid
in order to encompass both tidal flows and lee waves.
Their main result is that, by impacting the group velocity of the primary and secondary
waves, the mean advection velocity modifies the most unstable triads. They have predicted
and confirmed numerically that a strong enough advective flow enhances the instability of
the central branch (leading to large scale mode since one secondary wavelength is larger
than the primary one) with respect to the external branch. However, the model is not able
to explain the existence of an interesting stable region, at intermediate velocity in their
numerical simulations. To go beyond, it would be necessary to take into account the spatial
www.annualreviews.org • Instabilities of internal wave beams 13
growth of the secondary waves within the internal wave beam. Such a theory relying on
the extension of the classical absolute or convective instability is still to be derived.
3.7. Effect of the rotation
3.7.1. Theoretical study. When one includes Coriolis effects due to Earth’s rotation at a
rate ΩC , the dispersion relation of internal waves is modified and this has of course conse-
quences on the group velocity, which we showed to be intimately associated to the stability
of internal wave beams. Bordes et al. (2012a) have reported experimental signatures of
Triadic Resonant Instability of inertial gravity beams in homogenous rotating fluid. It is
thus expected that TRI will also show up when considering stratified rotating fluid.
Assuming invariance in the transverse y direction, the flow field may be written as
(ux, uy, uz) = (∂zψ, uy,−∂xψ) with ψ(x, z, t) the streamfunction of the non-divergent flow in
the vertical plane, and uy(x, z, t) the transverse velocity. Introducing the Coriolis parameter
f = 2ΩC sinβ, where β is the latitude, the dynamics of the flow field is given by the following
system of three equations
∂tb+ J(b, ψ)−N2∂xψ = 0, (21)
∂t∇2ψ + J(∇2ψ,ψ)− f∂zuy = −∂xb+ ν∇4ψ, (22)
∂tuy + J(uy, ψ) + f∂zψ = ν∇2uy. (23)
in which the equation for the buoyancy perturbation is not modified, while Equation (4)
has been modified and is now coupled to the dynamics of the transverse velocity uy.
As previously, one can study beams of general spatial profile, corresponding to the
superposition of time-harmonic plane waves with a dispersion relation (8) modified into
ω2 = N2 sin2 θ + f2 cos2 θ. The next step is again to look for subharmonic perturbations
in the form of fine-scale (with respect to the width of the beam), nearly monochromatic
wavepackets with frequency close to half the primary frequency. It is straightforward to see
that subharmonic waves propagate with an inclination φ given by sinφ = ((ω20/4−f2)/(N2−
f2))1/2 that vanishes when ω0/2 = f , i.e. at the critical latitude β ' 28.8 (MacKinnon
& Winters 2005). The modulus of the group velocity of subharmonic waves cg,± = (N2 −f2) sin(2φ)/(ω0k±) will thus also vanish at this latitude. The rotation reducing dramatically
the ability of subharmonic waves to escape, it may seriously reinforce the instability.
Karimi & Akylas (2017) have shown that it is possible to reproduce the asymptotic
analysis of the Triadic Resonant Instability with the inclusion of Earth’s rotation (see also
Karimi (2015)). One ends up with an unchanged equation (20) for the dynamics of the
primary wave, while the coupled dynamics of subharmonic waves is modified as follows
∂Ψ±∂τ
= −cg,± · ek0∂Ψ±∂η
+i
2
3f
κ2N
∂2Ψ±∂η2
− νκ2Ψ± + iδκ2
∣∣∣∣∂Ψ0
∂η
∣∣∣∣2 ψ± − γ ∂2Ψ0
∂η2Ψ∗∓, (24)
with δ and γ two parameters depending on the Coriolis parameter f , which vanish when
f tends to zero. The important modification is the appearance, on the right-hand side,
of the second linear term due to dispersion. It is important here because the first one
may disappear since the projection of the respective group velocity cg,± of subharmonic
envelopes Ψ± on the across beam direction ek0 may vanish.
Karimi & Akylas (2017) consider first weakly nonlinear sinusoidal wavetrains, empha-
sizing two interesting limits: the case far from the critical latitude allows one to recover the
14 Dauxois et al.
results of section 3.5 in which there is no preferred wavelength of instability in the inviscid
limit. On the other hand, when the group velocity of perturbations vanishes at the critical
latitude, energy transport is due solely to second-order dispersion. This process of energy
transport leads to the selection of a preferred wavenumber, independent of damping effects,
which may suppress the instability for a sufficiently large damping factor, permitting the
underlying wave to survive the instability. They obtained an expression for the growth rate
identical to the result in the inviscid limit of Young, Tsang & Balmforth (2008). It explains
that, at the critical latitude, additional physical factors, such as scale-selective dissipation,
must become important. This result has been shown numerically by Hazewinkel & Winters
(2011) in agreement with in situ measurements (Alford et al. 2007).
For beams, there is always a competition between energy extraction from the beam,
which varies with beam profile, and the proximity to the critical latitude, without forget-
ting the viscous effects on the fine-scale structure of disturbances. Relying on numerical
computations, it is possible to predict the stability properties for a given profile. In gen-
eral, it turns out that rotation plays a significant role in dictating energy transfer from an
internal wave to fine-scale disturbances via TRI under resonant configurations.
3.7.2. Experimental study. Until now, the only laboratory experiment that studied the
influence of rotation on the triadic instability of inertia-gravity waves, in a rotating stratified
fluid, was performed recently by Maurer, Joubaud & Odier (2016). In this study, the set-up
that was used by Bourget et al. (2013, 2014) was placed on a rotating platform, with a
range of rotation rates from 0 to 2.16 rpm, allowing the dimensionless Coriolis parameter,
f/N , to vary in a range from 0 to 0.45.
One of their main findings is the observation that the TRI threshold in frequency is
lowered (by about 20%), compared to the non-rotating case. An extension of the energy
approach developed in section 3.4 to the rotating case confirms this observation, by showing
that the finite-size effect of the beam width is reduced when rotation increases. This
enhancement of TRI only applies to a limited range of rotation rate since, when the rotation
rate overcomes half the primary wave frequency, TRI is forbidden due to the dispersion
relation not allowing the existence of the lowest-frequency secondary wave. The competition
between this limit and the finite-size effect reduction by rotation results in a minimum value
for the frequency threshold. The position of this minimum, observed around f/ω0 ' 0.35 in
the experiment, depends on the Reynolds number, defined as Re = Ψ0/ν. The transposition
of this result to high Reynolds number situations like in the ocean shows that the TRI
enhancement is then localized in a narrow Coriolis parameter range, with f = ω0/2, thus
recovering the critical latitude phenomenon.
When global rotation is applied to the fluid, another interesting feature is that it creates
an amplitude threshold for TRI. Indeed, it was discussed in section 3.2.3 that in the absence
of rotation, plane waves were always unstable. However, as shown in this section, the
instability at very low amplitude occurs when k+ tends towards k0, which implies ω+ →ω0 and therefore ω− → 0. But when there is rotation in the system, a zero-frequency
subharmonic wave is no longer allowed, hence the appearance of a threshold in amplitude,
which increases with f/N .
www.annualreviews.org • Instabilities of internal wave beams 15
4. STREAMING INSTABILITY
4.1. Introduction
Another important mechanism for the instability of internal gravity waves beams is the
generation of a mean flow, also called streaming instability2 by Kataoka & Akylas (2015).
Lighthill (1978a) and Andrews & McIntyre (1978) noticed in early times that internal
gravity wave beams share several properties with acoustic wave beams. In particular, both
kinds of waves may be subject to streaming in the presence of dissipative effects. Streaming
refers here to the emergence of a slowly evolving, non-oscillating, Eulerian flow forced by
nonlinear interactions of the oscillating wave-beam with itself (Nyborg 1965; Lighthill
1978b). As reviewed in Riley (2001), it is now recognized that streaming occurs actually
in a variety of flow models; it remains an active field of research for both theoretical (Xie
& Vanneste 2014) and experimental (Squires & Quake 2013; Moudjed et al 2014; Chraibi
et al. 2011) points of view.
The fact that dissipative effects are required to generate irreversibly a mean flow through
the nonlinear interactions of a wave beam with itself can be thought of as a direct conse-
quence of “non-acceleration” arguments that came up in the geophysical fluid dynamics con-
text fifty years ago with important contributions from Charney & Drazin (1961); Eliassen
& Palm (1961); Andrews & McIntyre (1976), among others. Plumb (1977) used those
ideas to propose an idealized model for the quasi-biennal oscillation (QBO), together with
an experimental simulation of the phenomenon (Plumb & McEwan 1978). The oscillations
require more than one wave beam, but Plumb (1977) discussed first how a single wave
beam propagating in a vertical plane could generate a mean flow. He predicted the vertical
shape of this mean flow, emphasizing the important role played by the wave attenuation
through dissipative effects. The experiment by Plumb & McEwan (1978) may be thought
of as the first quantitative observation of streaming in stratified fluids.
Those examples correspond, however, to a very peculiar instance of streaming, with
no production of vertical vorticity. By contrast, most applications of acoustic streaming
since the earlier works of Eckart (1948) and Westervelt (1953) involve the production
of vorticity by an irrotational wave. As far as vortical flows are concerned, Lighthill
(1978a) noticed important analogies between acoustic waves and internal gravity waves:
in both cases, vortical flows and propagating waves are decoupled at a linear level in the
inviscid limit, and steady streaming results from viscous attenuation of the wave amplitude.
In particular, Lighthill (1978a) noticed that streaming could generate a flow with vertical
vorticity. However, experimental observation of the emergence of a vortical flow in stratified
fluids through this mechanism remained elusive until recently. While studying the internal
wave generation process via a tidal flow over 3D seamounts in a stratified fluid, King, Zhang
& Swinney (2009) observed a strong flow in the plane perpendicular to the oscillating tidal
flow. For low forcing, this flow was found to be proportional to the square of the forcing
amplitude. That led them to invoke nonlinear interactions, either between the internal wave
beam and itself, or between internal waves and the viscous boundary layer. The analysis was
not pursued further, and the sign of the vorticity generated, opposite to the one discussed
in next subsections, remains puzzling.
A few years later, studying the reflection of an internal wave beam on a sloping bot-
tom, Grisouard and his collaborators have also discovered this mean-flow generation in
2This should not be mixed-up with the mechanism for planetesimal formation in astrophysics.
16 Dauxois et al.
experiments (Grisouard 2010; Leclair et al. 2011; Grisouard et al. 2013). The basic
configuration was an uniform beam reflecting onto a simple slope in a uniformly stratified
fluid. As predicted (Dauxois & Young 1999; Gostiaux et al. 2006), the interaction between
the incident and reflected waves produced harmonic waves, thereby reducing the amplitude
of the reflected wave. However, more surprisingly, they found that the reflected wave was
nearly absent because a wave-induced mean flow appeared in the superposition region of
the incident and reflected waves, progressively growing in amplitude. Comparing two- and
three-dimensional numerical simulations, they showed that this mean flow is of dissipa-
tive origin3 and three-dimensional. Its presence totally modifies the two-dimensional view
considered in the literature for reflection of internal waves. Indeed, there has been many
interesting theoretical studies of internal gravity waves-mean flow interactions (Bretherton
1969; Lelong & Riley 1991; Tabaei & Akylas 2003), but none of them considered the effect
of dissipation in three dimensions.
The complete and theoretical understanding of the generation of a slowly evolving vor-
tical flow by an internal gravity wave beam was possible using an even simpler set-up that
we describe in the following section. Bordes et al. (2012b) reported observations of a
strong mean flow accompanying a time-harmonic internal gravity beam, freely propagating
in a tank significantly wider than the beam. We describe below in detail the experimental
set-up and the observations, together with two related theories by Bordes et al. (2012b)
and Kataoka & Akylas (2015), which describe well the experimental results, by providing
the spatial structure and temporal evolution of the mean flow and illuminating the mech-
anism of instability. Those approaches bear strong similarities with the result obtained by
Grisouard & Buhler (2012), who used generalized Lagrangian mean theory, in order to de-
scribe the emergence of a vortical flow in the presence of an oscillating flow of a barotropic
tide above topography variations.
4.2. Experimental Observations
Bordes et al. (2012b) have studied an internal gravity wave beam of limited lateral extent
propagating along a significantly wider stratified fluid tank. Previously, most experimental
studies that were using the same internal wave generator (Gostiaux et al. 2007; Mercier et
al 2010) were quasi-two-dimensional (beam and tank of equal width) and therefore without
significant transversal variations.
Figure 4a presents a schematic view of the experimental set-up in which one can
see the generator, the tank and the representation of the internal wave beam generated
(see Bordes (2012) for additional details). The direct inspection of the flow field shows
an unexpected and spontaneously generated pair of vortices, emphasized in Figure 4b by
the tracer particles dispersed in the tank to visualize the flow field using particle image
velocimetry. This structure is actually a consequence of the generation of a strong mean
flow. This experiment provides therefore an excellent set-up to carefully study the mean-
flow generation and to propose a theoretical understanding that explains the salient features
of the experimental observations.
These observations are summarized in Figure 5, which shows side and top views not
only of the generated internal wave beam but also of its associated mean flow. One sees that
3Note however that transient mean flows can be generated by inviscid motion in the wake of apropagating internal wave packet (Bretherton 1969; van den Bremer & Sutherland 2014).
www.annualreviews.org • Instabilities of internal wave beams 17
0 5 10 15 20 25 30
−15
−10
−5
0
5
10
15(a) (b)
O
cg
cϕ
k
ω
θ
u
Wavem
aker
x
x (cm)
y
y(cm)
z
Figure 4
(a) Schematic representation of the experimental set-up with the generator on the left of the tank.
(b) Top view of the particle flow in a horizontal plane at intermediate depth.
−1
0
1
10 20 30
y(cm)
(mm/s)
x (cm)
z(cm)
z(cm)
x (cm) x (cm) x (cm)
y(cm)
−10
40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
100
−10
100
010
2030
010
2030 (a)
(c)
(b)
(d)
(e)
(g)
(f)
(h)
Experiment
Side viewTheory
Top view
Experiment Theory
Figure 5
Experimental (a,c,e,g) and theoretical (b,d,f,h) horizontal velocity fields ux for the primary wave
(top plots, obtained by filtering (Mercier, Garnier & Dauxois 2008) the velocity field at the
forcing frequency) and the mean flow (bottom plots, obtained by low-pass filtering the velocityfield) as reported respectively in Bordes et al. (2012b) and Kataoka & Akylas (2015). The four
left panels present the side view, while the right ones show the top view. The wave generator is
represented in grey with its moving part in black.
the wave part of the flow is monochromatic, propagating at an angle θ and with an amplitude
varying slowly in space compared to the wavelength λ. These waves are accompanied by a
mean flow with a jet-like structure, in the direction of the horizontal propagation of waves,
together with a weak recirculation outside the wave beam. Initially produced inside the
wave beam, this dipolar structure corresponds to the spontaneously generated vortex shown
in Figure 4b. Moreover, the feedback of the mean-flow on the wave leads to a transverse
bending of wave beam crests that is apparent in Figures 5e and 5f.
18 Dauxois et al.
4.3. Analytical Descriptions
4.3.1. A preliminary multiple scale analysis. Taking advantage of the physical insights pro-
vided by the experiments, Bordes et al. (2012b) have proposed an approximate description
that uses a time-harmonic wave flow with a slowly varying amplitude in space. The prob-
lem contains two key non-dimensional numbers, the Froude number U/λN and the ratio
ν/λ2N between the wavelength λ and the attenuation length scale of the wave beam due to
viscosity, λ3N/ν (Mercier, Garnier & Dauxois 2008). For analytical convenience, they con-
sidered a distinguished limit with the small parameter ε = Fr1/3, together with the scaling
ν/λ2N = ε/λν where λν ∼ 1. As usual, the appropriate scaling in the small parameter ε
is deduced from a mix of physical intuition and analytical handling of the calculations.
In their case, they were looking for a regime with small nonlinearity and dissipation. In
terms of the velocity components (ux, uy, uz), the buoyancy b and the vertical vorticity
Ω = ∂xuy − ∂yux, the governing dimensionless equations in this three-dimensional setting
read now
∇H · uH = −∂zuz, (25)
∂tb+ ε3 (u · ∇b) + uz = 0, (26)
∂tΩ + ε3 (uH · ∇HΩ + (∇H · uH) Ω + ∂x (uz∂zuy)− ∂y (uz∂zux)) = ελ−1ν ∇2Ω, (27)
∇2∂ttuz+∇2Huz = ελ−1
ν ∇4∂tuz−ε3(∂t(∇2H (u · ∇uz)− ∂z∇H (u · ∇uH)
)+∇2
H (u · ∇b)),
(28)
in which the index H in uH and ∇H reduces to the horizontal velocity field, gradient or
Laplacian operator.
Introducing rescaled spatial and time coordinates, a multiple scale analysis is now at
hand. Looking for a flow field in perturbation series ur = u0r + εu1
r + o(ε) for r = x, y
or z with a priori u0y = 0 as suggested by the structure of the beam, together with the
vertical vorticity field Ω = ε2Ω2 + ε4Ω4 + o(ε4), a tedious but straightforward application
of the multiple scale framework (with xi = εix and ti = εit) gives then, to the first three
orders, the structure of the beam: the first order ε0 provides the expressions for u0x and u0
z,
the second order ε1 gives the expression for u1y and finally order ε2 shows that u0
z does not
depend on the slow timescale t2.
Nonlinear terms contribute a priori for the first time to order ε3, but one interestingly
finds again (see section 2.3) that they vanish to this order. To order ε4, one obtains that the
term independent of the slow time t0 vanishes and, thus, nonlinear terms do not induce a
mean flow to this order either. It is only to order ε5 that nonlinear terms directly contribute
to the mean-flow generation. The governing equation of the vortical flow induced by the
mean flow is then given in the original dimensional units by
∂tΩ =∂xy U2
(2 cos θ)2+ ν∇2Ω (29)
where the overline stands for the filtering over one period and U(x, y) is the amplitude of
the wave envelope.
Several conclusions can be directly inferred from this analysis:
i) As emphasized by the first term on the right-hand side, nonlinear terms are crucial
as a source of vertical vorticity. Note that one recovers that the amplitude of the mean flow
is proportional to the square of the wave amplitude as has been invoked from experimental
(King, Zhang & Swinney 2009) or theoretical (Buhler 2009) results.
www.annualreviews.org • Instabilities of internal wave beams 19
ii) The variations of the wave field in the y-direction (implying ∂y 6= 0) are necessary for
nonlinearities to be a source of vertical vorticity. This illuminates why three-dimensional
effects are crucial and therefore why no mean-flow generation was noticed in two dimen-
sions (Mercier et al 2010; Grisouard et al. 2013).
iii) Finally, the viscous attenuation of the wave field in the x-direction (implying ∂x 6= 0)
is also necessary to produce vertical vorticity. In actual experiments, variations of the
amplitude in the x-direction can also come from finite size effects, but are not sufficient to
generate a mean flow.
One drawback of the above approach, however, is that it does not describe the feedback
of the mean flow on the waves. For this reason, the approach becomes inconsistent at long
time in the far field region.
The above combined experimental and analytical proof of the key role played by viscous
attenuation and lateral variation of the wave beam amplitude in the generation of the ob-
served mean flow has therefore motivated a more careful asymptotic expansion by Kataoka
& Akylas (2015), taking into account the two-way coupling between waves and mean-flow.
This two-way coupling accounts for the horizontal bending of the wave mean-field in Bordes
et al. (2012b) experiments, as explained in section 4.3.3.
4.3.2. Stability to three-dimensional modulations. Initially, Kataoka & Akylas (2013) were
interested in three dimensional perturbations of internal wave beams. Specifically, they
studied the stability of uniform beams subject to oblique modulations which vary slowly
in the along-beam and the horizontal transverse directions. Results turned out to be fun-
damentally different from that of purely longitudinal modulations considered in Tabaei &
Akylas (2003). Because of the presence of transverse variations, a resonant interaction be-
comes possible between the primary beam and three dimensional perturbations. Moreover,
their analysis revealed that three-dimensional perturbations are accompanied by circulating
horizontal mean flows at large distances from the vicinity of the beam.
They studied the linear stability of uniform internal wave beams with confined spatial
profile by introducing infinitesimal disturbances to the basic state, in the form of normal
modes, not only in the along-beam direction ξ but also in the horizontal transverse di-
rection y (see Figure 1). They used an asymptotic approach, valid for long wavelength
perturbations relative to the beam thickness. The boundary conditions combined with
the matching conditions between the solution near and far from the beam ensure that the
primary-harmonic and mean-flow perturbations are confined in the cross-beam direction.
The analysis brings out the coupling of the primary-harmonic and mean-flow pertur-
bations to the underlying internal wave beam: the interaction of the primary-harmonic
perturbation with the beam induces a mean flow, which in turn feeds back to the primary
harmonic via interactions with the beam. Whether this primary-harmonic-mean flow in-
teraction mechanism can extract energy from the basic beam, causing instability, depends
upon finding modes which remain confined in the cross-beam direction.
4.3.3. Complete model for the 3d propagation of small-amplitude internal wave beams.
In a second stage, Kataoka & Akylas (2015) have derived a complete matched asymptotic
analysis of the experiment performed by Bordes et al. (2012b) for a 3D Boussinesq weakly
nonlinear viscous fluid uniformly stratified. From their prior experience (Tabaei & Akylas
2003; Kataoka & Akylas 2013), they have chosen stretched along-beam spatial coordinate as
Ξ = ε2ξ, slow time as T = ε2t and transverse variations as Y = εy so that along-beam and
20 Dauxois et al.
transverse dispersions are comparable, together with variations in the cross-beam direction η
(see Figure 1). Combining this choice with a small nonlinearity scaling as ε1/2 and a weak
viscous dissipation νε2 that carry equal weight, they were able to fully analyze the mean
flow, separately near and far from the beam, before matching both solutions.
They derived a closed system of two coupled equations linking the amplitude of the
primary time harmonic U and the mean-flow component V∞ of the cross-beam velocity
field. The latter appears to be necessary for matching with the mean flow far from the
beam. The equation governing the dynamics of the mean flow reads
∂TV∞ = cos θ ∂YH(∫ +∞
−∞dη U∗
(∂U∂Ξ
+cot θ
2
∫ η
dη∂2U∂Y 2
)), (30)
where H(.) stands for the Hilbert transform in the transverse coordinate Y . This imme-
diately shows that transverse variations (∂Y 6= 0) of the beam are essential for having a
nonzero source term.
Since the generated mean vertical vorticity is given at leading order by Ω = cos θ ∂Y U =
(cos2 θ/ sin θ) ∂Y V∞+O(ε1/2), a direct comparison with Equation (29) is possible. The first
term in Equation (30), which involves derivatives in both horizontal coordinates corresponds
to the term identified by Bordes et al. (2012b), while this more complete analysis sheds
light on an additional term deriving from purely transverse variations.
Using an intermediate equation, Kataoka & Akylas (2015) end finally with the alter-
native and more elegant form
∂TV∞ = i∂YH(∫ +∞
−∞dη[(U∗∂ηU)T + ν∂ηU∗∂ηηU
]). (31)
Moreover, they show that to match inner and outer solutions, this induced mean flow turns
out to be purely horizontal to leading order and also dominant over the other harmonics.
The comparison of this theoretical description agrees very well with the experimental results
as beautifully emphasized by the different panels presented in Figure 5.
As far as a comparison with the experimental results of Bordes et al. (2012b) is
concerned, a common caveat of the predictions by Bordes et al. (2012b) and Kataoka &
Akylas (2015) is the assumption of small wavelength compared to the length scale of the
wave envelope, which is only marginally satisfied in the experiments. One may for instance
wonder if the horizontal structure of the observed waves is primarily due to the feedback of
the mean flow on the wave, or to the sole diffraction pattern of the wave due to this absence
of scale separation. This needs to be addressed in future works.
4.4. Forcing of Oceanic Mean Flows
Using an analysis based on the Generalized-Lagrangian-Mean (GLM) theory, Grisouard &
Buhler (2012) have also studied the role of dissipating oceanic internal tides in forcing
mean flows. For analytical convenience, they model wave dissipation as a linear damping
term −γbb in the buoyancy equation (3), and neglect the viscous term in the momentum
equation (2).
Within this framework, they discuss in detail the range of situations in which a strong,
secularly growing mean-flow response can be expected. Their principal results include the
derivation of an expression for the effective mean force exerted by small-amplitude internal
tides on oceanic mean flows. At leading order, taking into account the background rotation
www.annualreviews.org • Instabilities of internal wave beams 21
and using a perturbation series in small wave amplitude, they derive the following explicit
expression
∂tΩ +γbf
N2∂zb =
−iγbN2
2(ω2 + γ2b )ω
(∇u∗z ×∇uz) · ez, (32)
for the average over the tidal period of the vertical vorticity. It is remarkable that one
recovers in the presence of rotation a forcing term on the right-hand side that is analogous
to the forcing terms obtained by Bordes et al. (2012b) and Kataoka & Akylas (2015) in the
non-rotating case. In inviscid rotating flows, vortical modes are at geostrophic equilibrium,
and there is a frequency gap separating those geostrophic modes from inertia-gravity waves.
This frequency gap generally precludes interactions between geostrophic modes and wave
modes. The work of Grisouard & Buhler (2012), however, shows that the combination of
nonlinear and dissipative effects allows for a one-way energy transfer from inertia-gravity
wave modes to geostrophic modes, through a genuinely three dimensional mechanism. Using
Equation (32), Grisouard & Buhler (2012) compute the effective mean force numerically
in a number of idealized examples with simple topographies.
Although a complete formulation with dissipative terms in the momentum equation is
necessary, the conclusion of this important work by Grisouard & Buhler (2012) is that
energy of inertia-gravity waves in rotating fluids can be transferred to a horizontal mean
flow by a similar resonance mechanism as described in the experiment by Bordes et al.
(2012b). One understands therefore that mean flows can be generated in regions of wave
dissipation, and not necessarily near the topographic wave source.
5. CONCLUSIONS AND FUTURE DIRECTIONS
We have presented several recent experimental and theoretical works that have renewed the
interest of internal wave beams. After emphasizing the reason for their ubiquity in stratified
fluids – they are solutions of the nonlinear governing equations – this review has presented
the two main mechanisms of instability for those beams:
i) Triadic Resonant Instability. We have shown that this instability produces a direct
transfer of energy from large scales (primary waves) to smaller scales (subharmonic ones)
for inviscid plane waves, but that it is no longer true for internal wave beams since the
most unstable triad may combine subharmonic waves with larger and smaller wavelength.
Moreover, the effects of the finite size and envelope shape for the onset of Triadic Resonant
Instability have been overlooked. These features have to be taken into account to safely
reproduce the complete nonlinear transfer of energy between scales in the ocean interior or
in experimental analog (Scolan, Ermanyuk & Dauxois 2013; Brouzet et al. 2016a), and
therefore to find its stationary state, the so-called Garrett and Munk Spectrum (Garrett
& Munk 1975) or its possible theoretical analog, the Zakharov spectrum for the wave
turbulence theory (Nazarenko 2011).
ii) Streaming Instability. Now that the mechanism underlying streaming instability and
the conditions for its occurrence have been identified, several other examples will probably
be reported in the coming years. For example, such a mean-flow generation has also been
observed in a recent experiment (Brouzet et al. 2016b) for which the reflection of internal
gravity waves in closed domains lead to an internal wave attractor. Two lateral Stokes
boundary layers generate indeed a fully three-dimensional interior velocity field that pro-
vides the condition for the mean flow to appear. With a perturbation approach, Beckebanze
& Maas (2016) confirmed this theoretically and showed that the generated 3D velocity field
22 Dauxois et al.
damps the wave beam at high wave numbers, thereby providing a new mechanism to es-
tablish an energetic balance for steady state wave attractors. Semin et al. (2016) have
also recently studied experimentally the generation of a mean flow by a progressive internal
gravity wave in a simple two-dimensional geometry, revisiting an experimental analog of
the quasi-biennial oscillation (Plumb & McEwan 1978). They study the feedback of the
mean flow on the wave, an essential ingredient of the quasi-biennial oscillation.
Which is the dominant mechanism? Kataoka & Akylas (2016) have recently suggested
that streaming instability are central to three-dimensional internal gravity wave beam dy-
namics in contrast to the TRI of sinusoidal wave train relevant to uniform beams, the special
case of a internal plane wave with confined spatial profile. This review reinforces therefore
the need for more three-dimensional experiments studying wave-induced mean flow. In
particular, the conditions that favor mean-flow generation with respect to triadic resonant
interaction remains largely unknown. Angles of propagation? Three-dimensionality? Strat-
ification? This is an important question that needs to be addressed.
SUMMARY POINTS
In an incompressible non-rotating linearly stratified Boussinesq fluid,
1. Plane waves are solutions of the linear and nonlinear equations for any amplitude.
2. Internal wave beams, which correspond to the superposition of plane waves with
wave vectors of different magnitude but pointing in the same direction, are solutions
of the linear and nonlinear equations.
3. Plane waves solutions are always unstable by TRI.
4. General localized internal wave beams are stable while (quasi) spatial-harmonic
internal wave beams are unstable if the beam is wide enough.
5. In presence of rotation, beams of general spatial profile are more vulnerable to TRI
especially close to the critical latitude where nearly-stationary wavepackets remain
in the interaction region for extended durations, facilitating energy transfer.
6. Internal gravity wave beams with confined spatial profile are linearly unstable to
three-dimensional modulations.
7. When the wave beam is attenuated along its direction of propagation and when the
wave-envelop varies in the transverse horizontal direction, nonlinear interactions of
the wave beam with itself induce the emergence of a horizontal mean-flow with
vertical vorticity.
DISCLOSURE STATEMENT
The authors are not aware of any biases that might be perceived as affecting the objectivity
of this review.
ACKNOWLEDGMENTS
This work was supported by the LABEX iMUST (ANR-10-LABX-0064) of Universite de
Lyon, within the program ”Investissements d’Avenir” (ANR-11-IDEX-0007) operated by
the French National Research Agency (ANR). This work has achieved thanks to the re-
sources of PSMN from ENS de Lyon. We acknowledge the contributions of G. Bordes,
www.annualreviews.org • Instabilities of internal wave beams 23
B. Bourget, C. Brouzet, P.-P. Cortet, E. Ermanyuk, M. Le Bars, P. Maurer, F. Moisy, J.
Munroe, H. Scolan, and A. Wienkers to our research on this topic. We thank T. Akylas, V.
Botton, N. Grisouard, H. Karimi, T. Kataoka, L. Maas, T. Peacock, C. Staquet, B. Voisin
for helpful discussions.
LITERATURE CITED
Alexander M. 2003. Encyclopedia of the Atmospheric Sciences. Chapter Parametrization of Physical
Process: Gravity wave Momentum Flux, pp. 1669–1705. London, Academic/Elsevier.
Alford M., Peacock T, MacKinnon JA, Nash JD, Buijsman MC, Centuroni LR, Chao SH, Chang
MH, Farmer DM, Fringer OB, Fu KH, Gallacher PC, Graber HC, Helfrich KH, Jachec SM,
Jackson CR, Klymak JM, Ko DS, Jan S, Johnston TMS, Legg S, Lee IH, Lien RC, Mercier MJ,
Moum JN, Musgrave R., Park JH, Pickering AI, Pinkel R, Rainville L, Ramp ST, Rudnick DL,
Sarkar S, Scotti A, Simmons HL, St Laurent LC, Venayagamoorthy SK, Wang YH, Wang J, Yang
YJ, Paluszkiewicz T, Tang TY. (2015). The formation and fate of internal waves in the South
China Sea. Nature 521:65-69
Alford M., MacKinnon JA, Zhao Z., Pinkel R., Klymak J., Peacock T. 2007. Internal waves across
the Pacific. Geophys. Res. Lett. 34:L24601
Andrews DG, McIntyre ME. 1978. An exact theory of nonlinear waves on a lagrangian-mean flow.
J. Fluid Mech. 89:609–646
Andrews DG, McIntyre ME. 1976. Planetary waves in horizontal and vertical shear: The generalized
Eliassen-Palm relation and the mean zonal acceleration. J. Atmos. Sci. 33:2031–2048
Beckebanze F, Maas LRM. 2016. Damping of 3D internal wave attractors by lateral walls Proceedings
of the VIII International Symposium on Stratified Flows, San Diego, 29 August-1 September
2016, Editor: University of California at San Diego.
Bell TH. 1975. Lee waves in stratified flows with simple harmonic time dependence. J. Fluid Mech.
67:705–722
Benielli D, Sommeria J. 1998. Excitation and breaking of internal gravity waves by parametric
instability. Journal of Fluid Mechanics 374:117–144
Bordes G. 2012. Interactions non-lineaires d’ondes et tourbillons en milieu stratifie ou tournant.
Ph.D. dissertation, ENS de Lyon (https://tel.archives-ouvertes.fr/tel-00733175/en)
Bordes G, Moisy F, Dauxois T, Cortet PP. 2012a. Experimental evidence of a triadic resonance of
plane inertial waves in a rotating fluid Phys. Fluids 24:014105
Bordes G, Venaille A, Joubaud S, Odier P, Dauxois. 2012b. Experimental observation of a strong
mean flow induced by internal gravity waves. Phys. Fluids 24:086602
Bourget B. 2014. Ondes internes, de l’instabilite au melange. Approche exprimentale.. Ph.D. dis-
sertation, ENS de Lyon (https://tel.archives-ouvertes.fr/tel-01073663/en)
Bourget B, Dauxois T, Joubaud S, Odier P. 2013. Experimental study of parametric subharmonic
instability for internal plane waves. J. Fluid Mech. 723:1–20
Bourget B, Scolan H, Dauxois T, Le Bars M, Odier P, Joubaud, S. 2014. Finite-size effects in
parametric subharmonic instability. J. Fluid Mech. 759:739–750
Bretherton FP. 1969. On the mean motion induced by internal gravity waves. J. Fluid Mech.
36:785–803
Brouzet C, Ermanyuk EV, Joubaud S, Sibgatullin IN, Dauxois T. 2016a Energy cascade in internal
wave attractors. Europhysics Letters 113:44001
Brouzet C, Sibgatullin IN, Scolan H, Ermanyuk EV, Dauxois T. 2016b Internal wave attractors
examined using laboratory experiments and 3D numerical simulations. J. Fluid Mech. 793:109–
131
Buhler OK. 2009. Waves and Mean Flows. London, UK: Cambridge University Press
Callies J, Ferrari R, & Buhler O. 2014. Transition from geostrophic turbulence to inertiagravity
waves in the atmospheric energy spectrum. Proceedings of the National Academy of Sciences
24 Dauxois et al.
111:17033–17038
Charney JG, Drazin PG. 1961. Propagation of planetary-scale disturbances from the lower into the
upper atmosphere. J. Geophys. Res. 66:83–109
Chraibi H, Wunenburger R, Lasseux D, Petit J, Delville JP. Eddies and interface deformations
induced by optical streaming. J. Fluid Mech. 688:195–218
Clark HA, Sutherland BR. 2010. Generation, propagation, and breaking of an internal wave beam.
Phys. Fluids 22, 076601
Cole ST, Rudnick DL, Hodges BA, Martin JP. 2009. Observations of tidal internal wave beams at
Kauai Channel, Hawaii. J. Phys. Oceanogr. 39:421–436
Craik AD. 1988. Wave Interactions and Fluid Flows. London, UK: Cambridge University Press
Dauxois T, Young WR. 1999. Near critical reflection of internal waves. J. Fluid Mech. 390:271–295
Davis RE, Acrivos A. 1967. The stability of oscillatory internal waves. J. Fluid Mech. 30:723–736
Eckart C. 1948. Vortices and streams caused by sound waves. Phys. Rev. 73:68-76.
Eliassen A., Palm E. 1961. On the transfer of energy in stationary mountain waves. Geofysiske
Publikasjoner22:1–23
Fritts, D. C., Alexander, M.J. 2003. Gravity wave dynamics and effects in the middle atmosphere.
Reviews of Geophysics 41:1003
Garrett C, Kunze E. 2007. Internal tide generation in deep ocean Annu. Rev. Fluid Mech. 39:57–87
Garrett C, Munk W. 1975. Space-time scales of internal waves: A progress report J. Geophys. Res.
80:291–297
Gayen B., Sarkar S. 2013. Degradation of an internal wave beam by parametric subharmonic insta-
bility in an upper ocean pycnocline. J. Geophys. Res. Oceans 118:4689–98
Gerkema T., Staquet C., Bouruet-Aubertot P. 2006. Decay of semi-diurnal internal-tide beams due
to subharmonic resonance. Geophys. Res. Lett. 33:L08604
Gill A. E. 1982. Atmosphere-ocean dynamics Int. Geophys. Series, Academic press 30
Gostiaux L, Dauxois T., Didelle H., Sommeria J., Viboud S. 2006. Quantitative laboratory obser-
vations of internal wave reflection on ascending slopes Phys. Fluids 18:056602
Gostiaux L, Dauxois T. 2007. Laboratory experiments on the generation of internal tidal beams
over steep slopes. Phys. Fluids 19:028102
Gostiaux L, Didelle H, Mercier S, Dauxois T. 2007. A novel internal waves generator. Experiments
in Fluids 42:123–130
Grisouard N. 2010. Reflexions et refractions non-lineaires dondes de gravitee internes. Ph.D. dis-
sertation, Universite de Grenoble (http://tel.archives-ouvertes.fr/tel-00540608/en/)
Grisouard N, Leclair M, Gostiaux L, Staquet C. 2007. Large scale energy transfer from an internal
gravity wave reflecting on a simple slope. Proc. IUTAM 8:119–128
Grisouard N, Buhler O. 2012. Forcing of oceanic mean flows by dissipating internal tides. J. Fluid
Mech. 708:250–278
Hasselman K. 1967. A criterion fo nonlinear wave instability. J. Fluid Mech. 30:737–739
Hazewinkel J, Winters KB. 2011. PSI on the Internal Tide on a β Plane: Flux Divergence and
Near-Inertial Wave Propagation. Journal of Physical Oceanography 41:1673–1682
Hibiya T, Nagasawa M, Niwa Y. 2002. Nonlinear energy transfer within the oceanic internal wave
spectrum at mid and high latitudes. J. Geophys. Res. 107:3207
Joubaud S, Munroe J, Odier P, Dauxois T. 2012. Experimental parametric subharmonic instability
in stratified fluids. Phys. Fluids 24:041703
Johnston TMS, Merrifield MA, Holloway PE. 2003. Internal tide scattering at the Line Islands
Ridge. J. Geophys. Res. 108:3365
Johnston TMS, Rudnick DL, Carter GS, Todd RE, Cole ST. 2011. Internal tidal beams and mixing
near Monterey Bay. J. Geophys. Res. 116:C03017
Karimi HH. 2015. Doctoral dissertation. Department of Mechanical Engineering, MIT
Karimi HH, Akylas TR. 2014 Parametric subharmonic instability of internal waves: locally confined
beams versus monochromatic wave trains. J. Fluid Mech. 757:381–402
www.annualreviews.org • Instabilities of internal wave beams 25
Karimi HH, Akylas TR. 2017 Near-inertial parametric subharmonic instability of internal gravity
wave beams. Physical Review Fluids submitted.
Kataoka T, Akylas TR. 2013 Stability of internal gravity wave beams to three-dimensional modu-
lations. J. Fluid Mech. 736:67–90
Kataoka T, Akylas TR. 2015 On three-dimensional internal gravity wave beams and induced large-
scale mean flows J. Fluid Mech. 769:621–634
Kataoka T, Akylas TR. 2016 Three-diensional instability of internal gravity beams. Proceedings of
the VIII International Symposium on Stratified Flows, San Diego, 29 August-1 September 2016,
Editor: University of California at San Diego.
Khatiwala S. 2003 Generation of internal tides in an ocean of finite depth: analytical and numerical
calculations. Deep-Sea Res. 50:3–21
King B, Zhang HP, Swinney HL. 2009. Tidal flow over three-dimensional topography in a stratified
fluid. Phys. Fluids 21:116601
Koudella CR, Staquet C. 2006. Instability mechanisms of a two-dimensional progressive internal
gravity wave. J. Fluid Mech. 548:165–196
Lamb KG. 2004 Nonlinear interaction among internal wave beams generated by tidal flow over
supercritical topography. Geophys. Res. Lett. 31:L09313
Leclair M, Grisouard N, Gostiaux L, Staquet C, Auclair F. 2011. Reflexion of a plane wave onto a
slope and wave-induced mean flow. Proceedings of the VII International Symposium on Stratified
Flows, Rome, 2226 August 2011, Editor: Sapienza Universita di Roma
Lelong M-P, Riley J. 1991. Internal wave-vortical mode interactions in strongly stratified flows. J.
Fluid Mech. 232:1:19
Lerisson G, Chomaz JM. 2017. Global stability of internal gravity wave. Phys. Rev. Fluids submit-
ted.
Liang Y, Zareei A, Alam MR. 2017. Inherently unstable internal gravity waves due to resonant
harmonic generation. J. Fluid Mech. 811:400–420
Lien R-C, Gregg MC. 2001. Observations of turbulence in a tidal beam and across a coastal ridge.
J. Geophys. Res. 106:4575–4591
Lighthill J. 1978a. Waves In Fluids. London, UK: Cambridge University Press
Lighthill J. 1978b. Acoustic streaming. J. Sound Vib.61:391–418
Lighthill J. 1996. Internal waves and related initial-value problems. Dyn. Atmos. Oceans 23:3–17
MacKinnon JA, Winters KB. 2005. Subtropical catastrophe: Significant loss of low-mode tidal
energy at 28.9. Geophys. Res. Lett. 43:5
MacKinnon JA, Alford MH, Sun O, Pinkel R, Zhao Z, Klymak J. 2013. Parametric subharmonic
instability of the internal tide at 29N. J. Phys. Oceanogr. 43:17–28
Mauge R, Gerkema T. 2008. Generation of weakly nonlinear nonhydrostatic internal tides over large
topography: a multi-modal approach. Nonlin. Processes Geophys. 15:233–244
Maurer P, Joubaud S, Odier P. 2016. Generation and stability of inertiagravity waves J. Fluid Mech
808:539–561
McEwan AD. 1971. Degeneration of resonantly excited standing internal gravity waves. J. Fluid
Mech. 50:431–448
McEwan AD. 1973. Interactions between internal gravity wave and their traumatic effect on a
continuous stratification. Boundary-Layer Meteorol. 5:159–175
McEwan AD, Plumb RA. 1977. Off-resonant amplification of finite internal wave packets. Dyn.
Atmos. Oceans 2:83–105
Mercier MJ, Garnier N, Dauxois T. 2008. Reflection and diffraction of internal waves analyzed with
the Hilbert transform. Physics of Fluids 20:0866015
Mercier MJ, Martinand D, Mathur M, Gostiaux L, Peacock T, Dauxois T. 2010. New wave gener-
ation J. Fluid Mech. 657:308–334
Moudjed B, Botton V, Henry D, Ben Hadid H, Garandet JP. 2014. Scaling and dimensional analysis
of acoustic streaming jets. Phys. Fluids 26:093602
26 Dauxois et al.
Mowbray DE, Rarity BS. 1967. A theoretical and experimental investigation of the phase configu-
ration of internal waves of small amplitude in a density stratified fluid. J. Fluid Mech. 28:1–16
Mied RP. 1976. The occurrence of parametric instabilities in finite-amplitude internal gravity waves.
J. Fluid Mech. 78:763–784
Nazarenko S. 2011. Wave Turbulence. Springer-Verlag, Berlin Heidelberg
Nyborg WL. 1965. Acoustic streaming. Phys. Acoustics 2:265
Pairaud I, Staquet C, Sommeria J, Mahdizadeh M. 2010. Generation of harmonics and sub-
harmonics from an internal tide in a uniformly stratified fluid: numerical and laboratory experi-
ments. In IUTAM Symposium on Turbulence in the Atmosphere and Oceans (ed. D. Dritschel),
vol. 28, pp. 51–62. Springer
Peacock T, Echeverri P, Balmforth NJ. 2008. An experimental investigation of internal tide gener-
ation by two-dimensional topography. J. Phys. Oceanogr. 38:235–242
Phillips OM. 1966. The Dynamics of the Upper Ocean. Cambridge University Press, New York
Plumb RA. 1977. The interaction of two internal waves with the mean flow: Implications for the
theory of the quasi-biennial oscillation. Journal of the Atmospheric Sciences 34:1847–1858
Plumb R, McEwan A. 1978. The instability of a forced standing wave in a viscous stratified fluid :
A laboratory analogue of the quasi-biennial oscillation. J. Atmos. Sci. 35:18271839
Rainville L, Pinkel R. 1978. Propagation of low-mode internal waves through the ocean. J. Phys.
Oceanogr. 36:1220
Riley N. 2001. Steady streaming. Annual Review of Fluid Mechanics33:43–65
Sarkar S, Scotti A. 2016. Turbulence During Generation of Internal Tides in the Deep Ocean and
Their Subsequent Propagation and Reflection. Annual Review of Fluid Mechanics49:1
Scolan H, Ermanyuk E, Dauxois T. 2013. Nonlinear fate of internal waves attractors. Physical
Review Letters. 110:234501
Semin B, Facchini G, Petrelis F, Fauve S. 2016. Generation of a mean flow by an internal wave
Physics of Fluids. 28:096601
Staquet C, Sommeria J. 2002. Internal gravity waves: from instabilities to turbulence. Annu. Rev.
Fluid Mech. 34:559–593
Squires, T.M. and Quake, S.R. 2013 Microfluidics: Fluid physics at the nanoliter scale. Reviews of
Modern Physics, 77:3-977
Sun O, Pinkel R. 2013. Subharmonic energy transfer from the semi-diurnal internal tide to near-
diurnal motions over Kaena Ridge, Hawa. Journal of Physical Oceanography 43:766–789
Sutherland BR. 2010 Internal Gravity Waves. London, UK: Cambridge University Press
Sutherland BR. 2013. The wave instability pathway to turbulence J. Fluid Mech. 724:1–4
Tabaei A, Akylas TR. 2003. Nonlinear internal gravity wave beams. J. Fluid Mech. 482:141–161
Tabaei A, Akylas TR, Lamb KG. 2005. Nonlinear effects in reflecting and colliding internal wave
beams. J. Fluid Mech. 526:217-243
Thomas NH, Stevenson TN. 1972. A similarity solution for viscous internal waves. J. Fluid Mech.
54:495506.
van den Bremer TS, Sutherland BR. 2014. The mean flow and long waves induced by two-
dimensional internal gravity wavepackets. Physics of Fluids 26:106601.
Voisin B. 2003. Limit states of internal wave beams. J. Fluid Mech. 496:243–293.
Westervelt PJ. 1953. The theory of steady rotational flow generated by a sound field. The Journal
of the Acoustical Society of America 25:60–67
Wienkers AF. 2015. A critical amplitude for finite size parametric subharmonic instability: numer-
ical simulations. Master Report. Berkeley University and ENS de Lyon.
Wunsch C., Ferrari R. (2004). Vertical mixing, energy, and the general circulation of the oceans.
Annu. Rev. Fluid Mech., 36:281-314.
Xie JH, Vanneste J. 2014. Boundary streaming with Navier boundary condition. Phys. Rev. E
89:063010
Young WR, Tsang Y-K, Balmforth NJ. 2008. Near-inertial parametric subharmonic instability. J.
www.annualreviews.org • Instabilities of internal wave beams 27
Fluid Mech. 607:25–49
Zhou Q, Diamessis PJ. 2013. Reflection of an internal gravity wave beam off a horizontal free-slip
surface. Phys. Fluids 25:036601
28 Dauxois et al.